France 2030 and the European AI doctrine: a sovereignty playbook.

From defensive to operational

For most of the LLM era — call it 2020 to 2023 — the European narrative on AI was defensive. The dominant frame was "regulate Big Tech and hope a champion emerges". Public investment was real but scattered; private investment was real but smaller-scale than its US or Chinese equivalents; the strategic conversation was shaped less by what Europe would build than by what it feared others would impose.

That narrative shifted in 2024 and accelerated through 2025. By the time the Paris AI Action Summit closed in February 2025, a different frame was taking hold: operational, opinionated, funded. Public actors moved from "we want a European AI champion" to "we are funding consortium-led, vertical, sovereign platforms in priority sectors". Private actors moved from "we are exploring AI" to "we are co-investing in platform infrastructure with public partners".

This article maps the doctrine that emerged, the instruments that fund it, and the position Gen4Travel occupies inside it. The intended reader is anyone — investor, public-affairs lead, builder, executive — trying to understand whether Europe's sovereign-AI agenda is real or rhetorical, and whether to commit accordingly.

France 2030 and the generative AI call

The France 2030 plan, operated through Bpifrance, is the largest single industrial transformation programme in French history — over €54 billion across forty priorities, of which artificial intelligence is among the most strategic. Within France 2030, the call that funded Gen4Travel — "Soutien au déploiement de l'IA générative dans l'économie" — has a specific shape worth understanding because it reflects the underlying doctrine.

The call funds consortium-led projects rather than single-company R&D. It demands TRL 8 within two years rather than open-ended research. It targets priority sectors rather than horizontal AI. The bet is explicit: build vertical, sovereign, sector-anchored AI platforms rather than chase horizontal foundation models, where the US and China have a structural lead.

Travel was singled out for two reasons. Strategic weight: the sector contributes over €200 billion to French GDP, employs millions, and is among the most exposed to platform capture by US-based intermediaries. Data sovereignty dependence: the value of an agentic platform in travel is bounded by the quality and granularity of the data it can access; in a sector dependent on Big-Tech-hosted booking aggregators, a sovereign alternative is not a luxury, it is an existential question for European operators' margins.

The Gen4Travel consortium — Accor, ADP, Amadeus, Capgemini, Docaposte, Digital New Deal, EONA-X — was selected against this brief. The consortium structure reflects the doctrine: no single member could have built the platform alone; together, they cover the full vertical (hospitality, airports, distribution, technology, identity, data fabric, ethics governance), and the resulting infrastructure is reusable beyond travel through what we call the Gen4All replicability principle.

The Paris AI Action Summit

Held in February 2025, the Paris AI Action Summit marked an inflection point in the European posture. Heads of state, industry leaders, civil society, and researchers converged on a position around three principles.

• AI for the common good. A counter-narrative to the framing of AI as either pure commercial product or pure existential risk; an emphasis on AI as infrastructure that should serve broad public utility — health, education, climate, social services — alongside private deployment.
• Sustainable AI. A genuine engagement with the energy and resource footprint of frontier-scale model training, and with the question of which training and deployment patterns are compatible with European climate commitments.
• Trustworthy AI. An operational, regulation-grounded version of "AI safety" — not abstract alignment debates, but concrete obligations around audit, transparency, human oversight, accuracy and cybersecurity.

Beyond the principles, the Summit produced two concrete signals. France announced major private-investment commitments associated with the summit; the Commission proposed the InvestAI initiative, mobilising €200 billion across the European AI ecosystem — compute infrastructure, vertical platforms, and ecosystem support. The political signal was unambiguous: Europe is not surrendering the AI layer to its allies, however benign those alliances are.

The EU AI Continent plan

Released in 2025 and refined through the year, the AI Continent plan organises European AI capacity around five pillars. It is worth understanding because it is the framework against which European builders are increasingly measured for public co-investment.

The five pillars deserve a brief unpacking, because the rhetoric of "AI capacity" obscures meaningfully different things.

Pillar 1 — AI Factories. Compute-anchored ecosystems giving SMEs and researchers access to high-performance computing, datasets, and shared services. The first wave of AI Factories was announced in 2024–2025, distributed across member states, with explicit mandates to support local AI development.

Pillar 2 — AI Gigafactories. Frontier-scale training infrastructure. The largest of the pillars in raw infrastructure spend; the most directly aimed at narrowing the compute gap with US and Chinese frontier labs. Concrete projects under this banner are mobilising tens of billions of euros each.

Pillar 3 — Apply AI strategy. Vertical adoption in priority sectors: health, manufacturing, energy, transport, public services. Funds and accelerates production deployment, in contrast to the more research-leaning earlier pillars. Gen4Travel sits here.

Pillar 4 — Data Union strategy. The infrastructure layer that vertical AI depends on: data spaces, data intermediaries, data altruism, sector-specific data fabrics. EONA-X is a flagship in this pillar; Gen4Travel composes on top of it.

Pillar 5 — AI in Science. Research infrastructure, partnerships, and the broader academic ecosystem. Less directly relevant to commercial deployment, but a strategic complement that ensures the talent pipeline.

The AI Champions and consortium logic

Across France and the EU, public actors have moved toward identifying and supporting "AI Champions" — coalitions or single companies positioned to anchor sector verticals. The shift from "find one champion" to "structure consortium champions" is recent and important.

The original European-champion narrative imagined a single Mistral-equivalent in each domain. This proved harder than expected: travel is not "won" by a single technology vendor, because the value chain is distributed across operators, distributors, identity, payment, ground transport, hospitality. No vendor can credibly anchor the vertical alone.

The consortium model that emerged accepts this distribution and uses it as a strategic asset. Gen4Travel is the result of that doctrine: Accor for hospitality, ADP for airports, Amadeus for distribution, Capgemini for technology integration, Docaposte for sovereign identity, Digital New Deal for ethics governance, EONA-X for the neutral data fabric. Each partner brings irreplaceable assets; together they cover the vertical.

This pattern is being replicated. Health, energy, manufacturing and public services consortia are forming under similar logic: a small number of major sector actors plus a tech integrator plus a sovereign identity provider plus a data-space operator. The Gen4All replicability principle that Gen4Travel codifies — that the platform architecture is portable across sectors — is the consortium pattern made formal.

The bet — and how to know if it is working

The strategic question for the next two years is not whether Europe can fund credible AI projects. It can, and it is. The question is whether the projects funded compound into a coherent stack:

• Sovereign compute (AI Factories and Gigafactories)
• Sovereign data spaces (EONA-X and its sector peers)
• Sovereign foundation and specialised models
• Sovereign vertical platforms (Gen4Travel and its analogues in health, energy, etc.)
• Sovereign identity rails (eIDAS 2.0 wallets)
• Sovereign payment rails (where applicable)

If those pieces compose, Europe has, for the first time, an end-to-end AI infrastructure that does not depend structurally on US-hosted layers. If they do not — if individual projects ship and hit local optima but never plug into each other — the sovereignty narrative remains rhetorical.

The signals to watch: do AI Factories serve as compute hosts for vertical platforms like Gen4Travel? Do data spaces interoperate, or do they fragment into incompatible silos? Does the EUDI Wallet roll out on schedule and gain real adoption, or does it slip and rely on workarounds? Do consortium platforms expose interfaces that other consortia can compose with, or do they harden into the same kind of closed ecosystems they were meant to replace?

Gen4Travel was designed to answer these questions affirmatively. Our platform consumes EU compute infrastructure where available; our data fabric (EONA-X) interoperates by design with adjacent data spaces under Gaia-X conventions; our identity layer is the EUDI Wallet via Docaposte; our APIs are open for adjacent consortia to compose with. We are betting that the stack composes — and engineering accordingly.